Automatic Annotation Augmentation Boosts Translation between Molecules and Natural Language

TL;DR

This paper presents LA$^3$, a framework that uses large language models to augment molecular annotations, significantly improving AI models for molecule translation and generation tasks in biological research.

Contribution

The paper introduces LA$^3$, a novel annotation augmentation method that enhances datasets and improves molecular translation models' performance.

Findings

LA$^3$ improves dataset quality with varied annotations.

LaMolT5 trained on augmented data outperforms state-of-the-art models.

Up to 301% performance improvement with LA$^3$ augmentation.

Abstract

Recent advancements in AI for biological research focus on integrating molecular data with natural language to accelerate drug discovery. However, the scarcity of high-quality annotations limits progress in this area. This paper introduces LA$^3$, a Language-based Automatic Annotation Augmentation framework that leverages large language models to augment existing datasets, thereby improving AI training. We demonstrate the effectiveness of LA$^3$ by creating an enhanced dataset, LaChEBI-20, where we systematically rewrite the annotations of molecules from an established dataset. These rewritten annotations preserve essential molecular information while providing more varied sentence structures and vocabulary. Using LaChEBI-20, we train LaMolT5 based on a benchmark architecture to learn the mapping between molecular representations and augmented annotations. Experimental results on text-based *de novo* molecule generation and molecule captioning demonstrate that LaMolT5 outperforms state-of-the-art models. Notably, incorporating LA$^3$ leads to improvements of up to 301% over the benchmark architecture. Furthermore, we validate the effectiveness of LA$^3$ notable applications in *image*, *text* and *graph* tasks, affirming its versatility and utility.